WP2 Interpretation and downscaling of climate model

In this WP, the output from RCMs and Earth System Models generated in WP1 and in the database of the EU project ENSEMBLES will be evaluated and interpreted with respect to hydrologically relevant variables and relationships. Further downscaling and bias compensation will be addressed when relevant.

Results and deliverables

DBS-tailoring (Distribution Based Scaling)
Precipitation scaling involves two steps (see Figure below): removal of spurious low intensities (blue->green)and adjustment of the frequency distribution (green->red) (a). In temperature scaling, the frequency distribution is adjusted (grey->red) (b). The scaling substantially improves the reproduction of both runoff and snow pack in the subsequent hydrological simulation (blue->red) (c).

Link to User manual on DBS-tailoring.


Evaluation of high-resolution RCM simulations
A lot of work in Hydroimpacts 2.0 has focused on high-resolution RCM simulations and their applicability for hydrological climate change impact assessment. For example, the impact of spatial resolution on the reproduction of local short-term precipitation extremes has been assessed by analyzing RCM results on resolutions down to 6 km (Figure 1).

As the highest short-term intensities are generally produced by very localised, convective-type rainfall events (e.g. thunderstorms), a high RCM resolution is required to capture the spatial variability. Another aspect concerns the model description of convective processes, which is generally highly simplified in today’s RCMs.

Figure 1. RCM grid over Stockholm with grid size from 50 to 6.25 km.

The impact of spatial resolution on short-term extremes in terms of the 10-year Intensity-Duration-Frequency (IDF) curves is illustrated in Figure 2. Although the observed intensity is consistently underestimated in the RCM simulations, the agreement is substantially improved, to an increasing degree with decreasing duration. This shows that high-resolution RCM simulations better capture the extreme small-scale variability generally found in local, short-term observations. This may indicate that the results from future RCM projections on high spatial resolution are more representative of the local changes to be expected than what is found using coarser RCM grids and larger spatial averages.

Figure 2. 10-year IDF curve for Stockholm based on local observations (OBS) and historical RCM simulations with grid size from 50 down to 6 km.